I am a 2nd year Ph.D. student at HKU, advised by Lingpeng Kong and Qi Liu.

I am particularly interested in building scalable methods that enable models to produce useful feedback, learn effectively from that feedback, and improve their reasoning and decision-making over time.

• Developing post-training methods for critique, reflection, and self-improving reasoning — CTRL, Polaris, OPSD, and ReMiss
• Improving alignment and evaluation for vision-language models — VLFeedback, VLRewardBench, and Reka Flash
• Exploring training recipes for diffusion language models — Dream and Dream-Coder

Previously, I obtained my Master’s degree at Shanghai Jiao Tong University, under the supervision of Shuai Li. I received my Bachelor’s degree from IEEE Honor Class, Shanghai Jiao Tong University, where I was fortunate to work with Junchi Yan.